Hey,

there has been a lot of work over the past decade to get better support for 
Hybrid Laptops, Laptops that ship with an integrated GPU that is used by 
default and a more powerful dedicated GPU that is used on-demand
some of this work includes:
- switcheroo-control to query GPU information[1]
- the PrefersNonDefaultGPU key added to the XDG Desktop Standard [2]
- GNOME and KDE adding facilities to give users the ability to use multiple 
GPUs on demand [3]

sadly this hasn't worked out perfectly for a few reasons specifically:
- the specificaton gives two different definitions of what the key should do
- Desktop developers went with the "a GPU other than the default" definition
- Application developers went with "a more powerful discrete GPU" definition

this has created the problem that applications, taking Blender as an example, 
have set this key in hopes of it starting with a more powerful GPU, as one 
would expect for 3D modeling software
and while it achieves that on the systems it was designed for, it falls apart 
when used on something like a Desktop where its common to have a dedicated GPU 
and an integrated GPU as part of the CPU.

While this isn't a big problem in some cases in others it can cause 
applications to completely malfunction making it unclear to users what is 
happening.
Taking Steam as an example, on my Laptop it crashes when started under the 
discrete GPU and while this is a more extreme example this could also occur 
with other applications.

With all that, I'm a bit unsure how to move forward with this
fixes have been contributed upstream[4][5][6][7] however this work has largely 
gone nowhere yet while end-users continue to be exposed to the broken behavior
some packages in Fedora have decided to remove this variable
There are at least twos Fedora packages that have removed the key themselves to 
address bug reports of the wrong GPU being used[8][9], other maintainers will 
have to check for themselves if they want to adopt this themselves, I'm 
currently approaching a bunch of deevelopers to convince them to remove the 
key, so it might not even be needed.

I hope this information helps someone and that one day Fedora will properly 
function on multi-GPU systems.

Jan

P.S. sorry if this is a bit rambly, this email has almost been two years in the 
making

[1] https://gitlab.freedesktop.org/hadess/switcheroo-control/
[2] 
https://specifications.freedesktop.org/desktop-entry-spec/latest/recognized-keys.html
[3] https://www.hadess.net/2016/10/dual-gpu-integration-in-gnome.html
[4] https://gitlab.freedesktop.org/hadess/switcheroo-control/-/merge_requests/69
[5] https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/3193
[6] https://invent.kde.org/frameworks/kio/-/merge_requests/1556
[7] https://github.com/linuxmint/xapp/pull/178
[8] 
https://src.fedoraproject.org/rpms/xonotic/c/eca22baae61b51566cbcbc202a2d59c640ddbb59?branch=rawhide
[9] 
https://src.fedoraproject.org/rpms/supertuxkart/c/8e143f6cd414e7f9559565f920a9b30b72cfc452?branch=rawhide
-- 
_______________________________________________
devel mailing list -- devel@lists.fedoraproject.org
To unsubscribe send an email to devel-le...@lists.fedoraproject.org
Fedora Code of Conduct: 
https://docs.fedoraproject.org/en-US/project/code-of-conduct/
List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines
List Archives: 
https://lists.fedoraproject.org/archives/list/devel@lists.fedoraproject.org
Do not reply to spam, report it: 
https://pagure.io/fedora-infrastructure/new_issue

Reply via email to